APPLICATION OF ALGORITHMS FOR ANOMALY DETECTION IN HEALTH-ENABLED SENSOR-CLOUD INFRASTRUCTURE

Abstract

Real-time patient monitoring and early disease diagnosis are two ways that the healthcare industry is benefiting from the integration of sensors and cloud technology. In order to detect changes in patient's health, a variety of non-invasive sensors are applied to the skin to monitor various physiological parameters. The collected data are then wirelessly communicated to the cloud data center.  However, the transmitted data are susceptible to several sources of interference called anomalies. Anomalies is when a sudden change occurs from the expected sensor data generated. This may be as a result of sensor faults, measurement faults, injection and alteration by malicious attackers. Therefore, this research tends to conduct a survey on existing algorithms or techniques used for the detection of anomalies in health-enabled sensor-cloud infrastructure.   The processes adopted by the algorithms were identified and discussed exhaustively. In addition, the simulation setup and programming languages adopted to implement and evaluate the existing algorithms, followed by the limitations of the algorithms, which may lead to future research directions are captured in this paper. The outcome of the research shows that machine learning algorithms were predominantly adopted for detecting anomalies with the support of clustering and classification processes. Furthermore, Visual Basic.Net simulation tool and Python programming language was mostly adopted for experimentation and evaluation of the existing techniques. Limitations such as overfitting, under-fitting, computation complexity (time and memory space), and missing data are hindering the optimal performance of existing algorithm, which needs to be addressed in future researches.